The EU AI Act in plain English for Irish SMEs

The EU AI Act applies lightly to most Irish small businesses: if you use ordinary AI tools, your duties are honesty with customers and training your staff, not paperwork.

Does the EU AI Act apply to my business?

Almost certainly yes, but probably not in the way you fear. The Act covers anyone who puts an AI system on the EU market and anyone who uses one at work. A café with a chatbot answering enquiries is a user (the Act says “deployer”).

What matters is that the Act sorts AI by risk, not by company size. Almost everything a small Irish business uses sits in the lowest bands, where the obligations are modest. The heavy requirements (conformity assessments, technical documentation, registration) land on high-risk systems and on the companies that build AI models, not on the shop using one.

This is general information, not legal advice. If you use AI around hiring, credit or health, talk to a solicitor who knows the area.

The EU AI Act explained simply: four risk bands

Every AI system falls into one of four buckets:

  • Unacceptable risk (banned). Social scoring, manipulative systems that exploit vulnerable people, emotion recognition in workplaces and schools, most real-time biometric identification in public. If you are reading this, you are not doing any of them.
  • High risk. AI used in recruitment and managing staff, access to education, credit scoring, essential public services, and some safety components in regulated products. This band carries real obligations: risk management, data governance, human oversight, logging and documentation.
  • Limited risk (transparency). Chatbots, AI that generates or alters content, deepfakes. The duty here is honesty: people should know they are dealing with a machine, or looking at material a machine made.
  • Minimal risk. Everything else. Spam filters, the AI in your accounting software, a tool that drafts a social post. No specific obligations.

A typical Cork or Munster business lives in the bottom two bands, with one exception.

The one place a small business becomes high risk: hiring

If you use AI to screen CVs, rank candidates, or make decisions about staff promotion, pay or task allocation, that is high risk under the Act.

Most small businesses buy this kind of tool rather than build it, and then the provider carries most of the compliance burden. Your side is still real: use the system as intended, keep a person genuinely in the loop (not a rubber stamp), keep the logs it produces, and tell candidates an automated system is involved. If you paste CVs into a general chatbot and ask it to rank them, you have in effect built your own screening tool, and you own more of the risk than you might think.

The simplest way to stay out of this band is to keep AI on the admin side of hiring (drafting the job ad, tidying notes from an interview you sat in on) and keep the judgement with a person.

AI Act compliance for SMEs: what to actually do

The practical version, in order. None of it needs a consultant.

  1. Write down what you use. One page listing every AI tool in the business, who uses it and what for. It takes twenty minutes and often turns up a free app someone signed up for alone.
  2. Check nothing sits in the banned or high-risk bands. Hiring and staff management are the ones to look at closely.
  3. Be open where you should be. If a customer is talking to a chatbot, make that obvious from the first message. Label AI-made images or text where a reader would assume a person made it. I go through where to draw that line in should you tell customers you use AI?
  4. Train the people using it, and note that you did. An hour on what the tools can and cannot do, what must never be pasted in, and who to ask when unsure covers it for most small teams. Training your team on AI in an hour has a running order you can use as it is.
  5. Put it in a short policy. Half a page naming the approved tools and the rules for client data is enough. See how to write an AI policy for a small business.
  6. Do not confuse this with GDPR. Client-data risk is mostly a data protection question, which I cover in is it safe to put client data into AI tools?

When do the rules bite, now the deadlines have moved?

For a small business using ordinary tools, the rules that apply to you are already live, and the 2026 changes did not remove them. The Act came into force in August 2024. The bans and the staff training duty applied from February 2025, and the transparency duties for chatbots and AI-generated content from August 2026.

The big change came in July 2026, when the EU’s “Digital Omnibus” amendment took effect. It pushed the main high-risk rules (the hiring and credit ones above) back from August 2026 to December 2027, and the rules for AI built into regulated products such as machinery and medical devices to August 2028. It also softened the wording of the training duty: you now have to take reasonable steps to build your team’s understanding of AI, rather than guarantee any particular level of it. The duty is still there, and doing nothing at all still falls short.

Who enforces the AI Act in Ireland?

Ireland has spread enforcement across regulators that already exist rather than creating one new AI watchdog. The Regulation of Artificial Intelligence Bill 2026, published by the Department of Enterprise, Tourism and Employment in June 2026, gives market surveillance roles to bodies such as the Data Protection Commission, the Central Bank and the Competition and Consumer Protection Commission,. A new AI Office of Ireland coordinates them and is the single point of contact with Brussels, but it does not police businesses itself.

In practice, the regulator you already deal with is the one most likely to ask about AI: the Central Bank for a broker, the Data Protection Commission wherever personal data is involved. The fines quoted in headlines (up to €35 million or 7% of global turnover) attach to the banned practices, and for small firms the Act caps fines at the lower of the fixed sum and the percentage. Proportionality for SMEs is written into the law.

The honest read: this is a housekeeping job, not an existential one. An afternoon listing your tools, checking hiring is clean, labelling your chatbot and briefing your staff puts you in a reasonable position.

If you would rather that afternoon produced a plan for using AI well as well as legally, that is roughly what the AI Assessment does: a fixed-fee look at how your week actually runs, what is worth automating and what to keep away from, with a follow-up call once you have read the report. If you are still deciding whether any of it is worth the bother, is AI actually worth it for a small business? is the better place to start.

Book the AI Assessment

The assessment is credited in full if you go on to a build, and if I can’t find you five hours a week you pay nothing. See how it works.